Some shortcomings of long-term working memory.

Within the framework of their long term working memory theory, Ericsson and Kintsch (1995) propose that experts rapidly store information in long-term memory through two mechanisms: elaboration of long-term memory patterns and schemas and use of retrieval structures. They use chess players' memory as one of their most compelling sources of empirical evidence. In this paper, I show that evidence from chess memory, far from supporting their theory, limits its generality. Evidence from other domains reviewed by Ericsson and Kintsch, such as medical expertise, is not as strong as claimed, and sometimes contradicts the theory outright. I argue that Ericsson and Kintsch's concept of retrieval structure conflates three different types of memory structures that possess quite different properties. One of these types of structures--generic, general purpose retrieval structures--has a narrower use than proposed by Ericsson and Kintsch: it applies only in domains where there is a conscious, deliberate intent by individuals to improve their memory. Other mechanisms, including specific retrieval structures, exist that permit a rapid encoding into long-term memory under other circumstances.